CN100361461C - Terminal to terminal running performance monitoring method based on sampling measurement - Google Patents
Terminal to terminal running performance monitoring method based on sampling measurement Download PDFInfo
- Publication number
- CN100361461C CN100361461C CNB2005100376650A CN200510037665A CN100361461C CN 100361461 C CN100361461 C CN 100361461C CN B2005100376650 A CNB2005100376650 A CN B2005100376650A CN 200510037665 A CN200510037665 A CN 200510037665A CN 100361461 C CN100361461 C CN 100361461C
- Authority
- CN
- China
- Prior art keywords
- point
- mask
- cooperative
- sampling
- result
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
Images
Landscapes
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The present invention discloses a method for monitoring end-to-end running performance based on sampling measurement. The running situation of the flow rate of a network is monitored by a method that monitoring points are arranged on the network to passively measure the network single delay and the packet loss rate, a cipher key for unionizing all the monitoring points is set, a hash value is generated by the cipher key and messages, a certain proportion of messages are randomly extracted on the network, the monitoring points are provided with two buffer pools, former 40 bytes of a head of the messages and 8 bytes of time stamps of the messages are stored in one current buffer pool, the other buffer pool is used for forwarding data, and the network single delay and the packet loss rate are obtained by data operation; therefore, the running performance of the network is obtained. The present invention has the advantages of smaller influence to performance of a measurer, high safety of a measuring method, and high sampling precision; furthermore, the same messages can be guaranteed to be measured by the different measuring points without sampling mask code negotiation, etc.
Description
Technical field
The present invention relates to a kind of method that is used to detect network performance, relate in particular to a kind of terminal to terminal running performance monitoring method based on sampling measurement.
Background technology
The theoretical research and the technological development of Internet management are not a new problem, but mainly concentrate on network management in the past.Yet, since middle nineteen nineties in last century, development along with Internet, large-scale company, government bodies, R﹠D institution, the private enterprise and ordinary consumer utilize the basic means of carrying out business activity based on the service conduct of Internet gradually, Service Management and user management become the emphasis of Internet management, the research emphasis of Internet Service Management also forwards acceptable service quality (quality of service, management QoS) to from the validity management.The current network management is used for network management and system management mostly, and these management mainly come supervising the network from the viewpoint of network and system, is not suitable for the management of present end to the end service.The network manager for management end to end service, need service level managment and Reporting Tools, the demand that no longer can satisfy the network manager based on fault detect and the fix tool of SNMP based on the Internet Service Management.
The current service system is provided by a plurality of service providers, so, in order to provide end to the user, just require each service provider can both guarantee the availability and the performance of its appropriate section to holding service quality to guarantee.The user is indifferent to each part that service is provided, and only is concerned about QoS.Expectation to QoS makes the user consult the QoS rank with their service provider, and this realizes by SLA (service level agreements).SLA be between service provider and the user for specifying the QoS rank of expectation, the service expectation behavior of signing and the contract of qos parameter.Main following three aspects of the research of sla management at present: SLA parameter-definition, the present public standard that does not also have the SLA parameter; SLA measures, and how main research accurately measures the QoS that the service provider offers the user, and the SLA measurement is the basis of sla management; The SLA/QoS management, monitoring in real time, analysis and report offer user's QoS.
Opposite end to the key of holding performance to monitor is unified monitoring content, makes the performance state of network not only vertically comparable, and has horizontal comparativity, could form the correct evaluation to the network behaviour in service.But, it seems that by present case the research of this respect does not also form unified standard.The IPPM working group of IETF has defined the parameter suggestion of a series of description network performances, as one-way delay, and one-way packet loss, round-trip delay etc., the clear and definite parameter of measuring object, unit, definition and method.And set up Current SurveyorNetwork Map at http://www.advanced.org/csg-ippm/ in view of the above, the target test packet is measured one-way delay and packet loss between 38 host nodes of the U.S. when having by transmission.The network condition report that the definition that this seemingly unique basis has been shaped is monitored network performance.With reference to the achievement of external job family, set up unified complex, the SLA parameter-definition is the primary prerequisite of network performance monitoring.
Because there is the multifrequency nature, user of very big heterogeneity, huge delivery flow rate, the flow demand to various different application in Internet, so network behavior research is quite challenging and quite complicated.Because the real network behavior can not be described and analyze to the prior theory model effectively, uses these models to be difficult to draw correct conclusion usually.Therefore be widely used based on the network management of measuring, there are CAIDA, MERIT, ITA, OC3 and IETF in main research institution.With after the data that are connected are gathered, state that need awareness network operation from past data to know the problem that may occur in advance, is rationally arranged offered load to the network equipment.Simultaneously, under abundant understanding equipment and the operating position that is connected, guidance is made in the network planning in future.How the work of both at home and abroad the state parameter sequence of network being analyzed does not at present also launch.The network service that at present has a lot of mechanisms that himself is provided is abroad carried out Measurement and analysis to a certain degree and is cooperated mutually, in the hope of obtaining full appreciation to network state, as CAIDA (Cooperative Association for Internet Data Analysis) to the measurement of network data flow and the work of analysis aspect.But research at present is to carry out according to the requirement of some application-specific mostly, and target is more random, does not also form than rounded system, and some theoretical models and the practice that have occurred also have a certain distance.By historical data the time-delay of network, the holding load of responding with in advance that throughput is carried out short-term are dynamically adjusted with some simple algorithm as University of California; The Boston University carries out the selection of dynamic state server by the test data to the available bandwidth of channel, but it just selects according to the present situation, and does not do more long-range analysis.Generally speaking, existing research is directly to carry out the instant optimization of some offered loads according to measured value, not to Long-term analysis, the integrated use of data.Therefore we can say that the state parameter of network is is systematically analyzed and researched also is in the starting stage at present.
Present measurement project mainly is that the invasion formula is measured, non-invasion formula is measured and polling mib.Invasion is measured many direct measuring method, but the invasion flow can influence network traffics; Non-invasion does not influence network traffics during measuring, but easy drop-out in the high speed network environment, and be difficult to the monitoring side to holding network performance; The polling mib method of measurement is simple, but some data necessary is difficult to obtain, and is difficult to use in the extensive high speed network environment.
At present main measurement project has: the skitter project survey of CAIDA from 22 source points to IP path, the whole world and RTT, to obtain and to follow the tracks of global Internet topology.The CoralReef project is that CAIDA uses passive flux monitor Collection and analysis data, and CoralReef can be used in the high speed backbone network environment of OC3, OC12 and OC48 now.The purpose of the NIMI project of NSF and DARPA is to set up network measure general-purpose platform architecture, and the NIMI prototype is followed IPPM and estimated standard, measures delay, packet loss and unidirectional route.The IEPM project is the end-to-end Internet performances of tissue use PingER instrument active monitoring such as SLAC/DOC.Abilene is the senior backbone network that the U.S. connects LAN Gbps exchange point, supports university to carry out the research work that high-level network is used, and the network monitoring behavior mainly is that broadcasting ping initiatively measures and based on the measurement of SNMP.Mantra is the instrument that CAIDA is used for monitoring at router level various multicast behaviors.The measurement of NLANR and operating analysis group (MOAT) are safeguarded a network analysis architecture (NAI), be used to study the Internet service model and estimate, the PMA project of MOAT is obtained the IP header information and is obtained SNMP ROM data from a plurality of measurement points from log file, the AMP project is obtained initiatively measurement data from more than 100 campus networks.Network weather service (NWS) is a distributed system, cycle monitoring and dynamic prediction NPACI network performance.
Traditional sla management instrument can not be collected performance parameter, promptly allows to collect performance parameter, the relation between the also difficult collaborative parameters, and whether the service of determining reaches the quality level of SLA agreement defined.The network management solution that the service provider is present can not realize end to holding performance management, and therefore, SLA impels the service provider to turn to the sla management instrument from traditional solution based on SNMP.The instrument that has some grades of service to follow the tracks of and report in the market, these instruments also need certain hour just from traditional network and system management divert service management but management system really can reach the SLA of ISP management.These instruments comprise: the Cisco Works2000 SMS instrument of Cisco company, the Pulsar xSP of Response Networks, the Info Vista system of Info Vista company, the VitalSuite of INS, the Network HealthReporter of ConCord Communications and the Firehunter of Agilent Technologies etc.The product that wherein has only can the monitor network service, and what have can the monitor terminal application program of user, but these instruments substantially all are to adopt the polling mib database and use simple ping instrument, and can only measure the SLA of low-speed small network.
The method of traditional measurement one-way latency and packet loss is the active method of measurement, be that active mode arrives measurement point B from measurement point A transmitted traffic, according to the unidirectional packet loss between information calculations measurement point A, the B of test packet and these two performance measurement indexs of one-way latency.But initiatively method of measurement has the shortcoming of two aspects: (1) active test stream amount is disturbed the ruuning situation of network normal discharge; (2) from measurement point A to the router the measurement point B with gateway to adopting different processing modes to different flow, may cause losing or queueing delay of test traffic, make measurement result can not reflect the performance condition of real network end-to-end like this.
Summary of the invention
The invention provides the fail safe and the sampling control precision of a kind of efficiency of measurement that can improve measurement point, measuring system and can reduce between the distributed measurement point cooperative information based on the terminal to terminal running performance monitoring method of sampling measurement.
The present invention adopts following technical scheme:
A kind of terminal to terminal running performance monitoring method based on sampling measurement:
The 1st step: on network, arrange at least 2 monitoring points, measuring appliance and server are set on each monitoring point, dispose the monitoring range of each monitoring point in each monitoring point, whole system can be measured the performance of n (n-1) paths at most, for a measurement point, for the measuring route of putting cooperative point from local measurement with from the measuring route that cooperative is put local measurement point two memory spaces are set all, a memory space is used to store intermediate object program packet_storage, another memory space is used to store the final measurement performance_storage in this path, at each monitoring point GP configuring S clock system, tell the light of 10% light intensity to measuring appliance by optical splitter from measuring optical fiber in each monitoring point
The 2nd step: identical key value key of configuration on each measurement point is provided with two Buffer Pools on measuring appliance: storage Buffer Pool basket1 and intransit buffering pond basket2, and the time granularity time_scale of performance processing,
The 3rd step: configuration sampling proportion ratio separately on each measurement point, sampling proportion ratio span is between 0 to 1, and a sampling mask lengths sets mask_length is set, and the ensemble space size is 16, initial value is empty, with sampling proportion ratio successively with 1/2
1, 1/2
2, 1/2
3, 1/2
4, 1/2
5, 1/2
6, 1/2
7, 1/2
8, 1/2
9, 1/2
10, 1/2
11, 1/2
12, 1/2
13, 1/2
14, 1/2
15, 1/2
16This 16 number compares, if ratio is greater than a certain number, then with first less than the denominator index recording of the number of sampling proportion ratio in sampling mask lengths sets mask_length, sampling proportion ratio deduct this first continue to compare less than the result of the number back gained of sampling proportion ratio with after this number, with in the number after this first less than the denominator index recording of gained result's number in sampling mask lengths sets mask_length, then the gained result is deducted above-mentioned first number less than the gained result, so repeatedly, till this 16 number is relatively finished
The 4th step: from sampling mask lengths sets mask_length, read element successively, define a Bit String, the length of this Bit String is the current value that reads element, with in the mask_length set less than being set to 0 on the current all elements value position of reading element value, be set to 1 on all the other positions, this Bit String as a mask, is recorded among the mask set mask
The 5th step: measuring appliance is received each message execution following steps:
5.1 in a single day measuring appliance receives that a message makes a call to a current time stamp of measuring appliance for this message immediately;
5.2 from heading, extract source IP, the place IP of 32 bits of 32 bits, the source port of 16 bits, the place port of 16 bits, the message identification IPID of 16 bits,
5.3 back 16 bits of source IP are carried out 3 left loop computations, and 16 bits carry out XOR behind result that will generate and the place IP, generate one 16 Bit String hash1; Preceding 16 bits of source IP are carried out 4 left loop computations, and the result that will generate and preceding 16 bits of place IP carry out XOR, generate one 16 Bit String hash2; 16 bits of source port are carried out 5 left loop computations, and 16 bits of result that will generate and place port carry out XOR, generate one 16 Bit String hash3; With above-mentioned 3 16 Bit String hash1, hash2 and hash3 and message identification IPID carry out XOR, generate the cryptographic Hash hash of 16 Bit Strings,
5.4 the cryptographic Hash hash and the key key that generate are carried out the final cryptographic Hash hash of XOR generation;
5.5 from mask set mask, read element successively, first bit extracts the Bit String of mask-length since the final hash value hash of 16 bits, and this Bit String and the current mask that reads compared, extraction has the message of the Bit String identical with mask, and 8 byte times of preceding 40 bytes of this heading and this message are stabbed be stored in the current Buffer Pool
The 6th step: continued to carry out the 5th step and handle newly arrived message, if current Buffer Pool is full, current Buffer Pool is set to the intransit buffering pond, another Buffer Pool in the measuring appliance is set to current Buffer Pool simultaneously, and the data in the intransit buffering pond are transmitted to background server in the mode of UDP message
The 7th step: background server is accepted the UDP message from measuring appliance, and reads heading details and time stab information from the UDP message,
The 8th step: according to the sourcesink Ip address of heading, and the address space of each measurement point that disposes in server administration, heading information and time stab information thereof are stored among the intermediate storage result space packet_storage in path under this message,
The 9th step: continue to handle the next message that arrives according to the 7th step, if Measuring Time enters a new time granularity interval, then heading information and the time stab information of putting in the last time granularity interval the intermediate storage result space packet_storage in local measurement point path from cooperative in the cooperative point thereof is sent in the local measurement point server, sampling proportion with cooperative point is sent to the local measurement point simultaneously
The 10th step: compare the sampling proportion of local measurement point and the sampling proportion of cooperative point, use little sampling proportion to continue to use 5.2 methods that went on foot for 5.5 steps to carry out secondary sample to the data in the last time grain size intervals of intermediate storage result space with big sampling proportion
The 11st step: cooperative is put in the local measurement point path in the last time grain size intervals of cooperative point in the intermediate storage result space number number1 of element deduct local measurement and put in the time grain size intervals number number2 of element in the intermediate storage result space, and with the two result divided by the number number1 of element in the intermediate storage result space in the last time grain size intervals of cooperative point, its result is exactly the unidirectional packet loss from the cooperative point to local measurement point path, and the result is stored among the final measurement memory space performance_storage in this path
The 12nd step: relatively cooperative is put in the local measurement point path in the last time grain size intervals of cooperative point the element in the intermediate storage result space and local measurement and is put the element in the intermediate storage result space in the time grain size intervals, select the source IP of element in two set, place IP, source port, the place port, the heading data that message identification IPID is identical, and the element timestamp that uses cooperative point deducts the element time stamp data of local measurement point, and the result is stored in the set of time difference
The 13rd step: from the time difference set, read the number of set element, the value of least member and the value of middle big or small element, value with least member is minimum one-way latency, value with the big or small element in centre is the intermediate value one-way latency, and with above-mentioned three values store into from the cooperative point to local measurement point path final measurement memory space performance_storage
The 14th step: the result of the number of minimum one-way latency, intermediate value one-way latency, unidirectional packet loss and measurement on will going up in the time grain size intervals from the cooperative point to local measurement point path is sent to the final measurement memory space performance_storage from the cooperative point to local measurement point path of cooperative point server.
2, the terminal to terminal running performance monitoring method based on sampling measurement according to claim 1 is characterized in that the time granularity time_scale that described performance of the 2nd step is handled is set to 5 minutes.
Compared with prior art, the present invention has following advantage:
(1) mainly is XOR, shift operation based on many masks of sign methods of sampling, is convenient to realize, use this method very little the performance impact of measuring appliance with hardware;
(2) ident value has been introduced key in handling, and on randomness that does not influence ident value and concertedness basis, has increased the fail safe of method of measurement;
(3) masks of measuring make sampling probability freely to adjust more, have the multiple choices sampling probability, realize that the sampling control precision arrives 1/65536;
(4) because the mask computational methods determine the sampling sample of big sampling probability to comprise little sampling probability sample, therefore a plurality of measurement points do not need to adopt identical sampling probability, do not need to carry out sampling mask yet and consult, can guarantee that different measurement points measures identical message.
Description of drawings
Fig. 1 is a network end-to-end performance passive measurement system assumption diagram;
Fig. 2 is measuring system network environment figure, wherein, measuring appliance is connected with server by a special circuit, is used to transmit measurement data and control information, and when needed, mutual by other equipment on server and Intranet (garden net) and the Internet (the Internet).
Embodiment
Embodiment 1
A kind of terminal to terminal running performance monitoring method based on sampling measurement:
The 1st step: on network, arrange at least 2 monitoring points, measuring appliance and server are set on each monitoring point, dispose the monitoring range of each monitoring point in each monitoring point, whole system can be measured the performance of n (n-1) paths at most, for a measurement point, for the measuring route of putting cooperative point from local measurement with from the measuring route that cooperative is put local measurement point two memory spaces are set all, a memory space is used to store intermediate object program packet_storage, another memory space is used to store the final measurement performance_storage in this path, at each monitoring point GP configuring S clock system, tell the light of 10% light intensity to measuring appliance by optical splitter from measuring optical fiber in each monitoring point
The 2nd step: identical key value key of configuration on each measurement point is provided with two Buffer Pools on measuring appliance: storage Buffer Pool basket1 and intransit buffering pond basket2, and the time granularity time_scale of performance processing,
The 3rd step: configuration sampling proportion ratio separately on each measurement point, sampling proportion ratio span is between 0 to 1, and a sampling mask lengths sets mask_length is set, and the ensemble space size is 16, initial value is empty, with sampling proportion ratio successively with 1/2
1.1/2
2, 1/2
3, 1/2
4, 1/2
5, 1/2
6, 1/2
7, 1/2
8, 1/2
9, 1/2
10, 1/2
11, 1/2
12, 1/2
13, 1/2
14, 1/2
15, 1/2
16This 16 number compares, if ratio is greater than a certain number, then with first less than the denominator index recording of the number of sampling proportion ratio in sampling mask lengths sets mask_length, sampling proportion ratio deduct this first continue to compare less than the result of the number back gained of sampling proportion ratio with after this number, with in the number after this first less than the denominator index recording of gained result's number in sampling mask lengths sets mask_length, then the gained result is deducted above-mentioned first number less than the gained result, so repeatedly, till this 16 number is relatively finished
The 4th step: from sampling mask lengths sets mask_length, read element successively, define a Bit String, the length of this Bit String is the current value that reads element, with in the mask_length set less than being set to 0 on the current all elements value position of reading element value, be set to 1 on all the other positions, this Bit String as a mask, is recorded among the mask set mask
The 5th step: measuring appliance is received each message execution following steps:
5.1 in a single day measuring appliance receives that a message makes a call to a current time stamp of measuring appliance for this message immediately;
5.2 from heading, extract source IP, the place IP of 32 bits of 32 bits, the source port of 16 bits, the place port of 16 bits, the message identification IPID of 16 bits,
5.3 back 16 bits of source IP are carried out 3 left loop computations, and 16 bits carry out XOR behind result that will generate and the place IP, generate one 16 Bit String hash1; Preceding 16 bits of source IP are carried out 4 left loop computations, and the result that will generate and preceding 16 bits of place IP carry out XOR, generate one 16 Bit String hash2; 16 bits of source port are carried out 5 left loop computations, and 16 bits of result that will generate and place port carry out XOR, generate one 16 Bit String hash3; With above-mentioned 3 16 Bit String hash1, hash2 and hash3 and message identification IPID carry out XOR, generate the cryptographic Hash hash of 16 Bit Strings,
5.4 the cryptographic Hash hash and the key key that generate are carried out the final cryptographic Hash hash of XOR generation;
5.5 from mask set mask, read element successively, first bit extracts the Bit String of mask-length since the final hash value hash of 16 bits, and this Bit String and the current mask that reads compared, extraction has the message of the Bit String identical with mask, and 8 byte times of preceding 40 bytes of this heading and this message are stabbed be stored in the current Buffer Pool
The 6th step: continued to carry out the 5th step and handle newly arrived message, if current Buffer Pool is full, current Buffer Pool is set to the intransit buffering pond, another Buffer Pool in the measuring appliance is set to current Buffer Pool simultaneously, and the data in the intransit buffering pond are transmitted to background server in the mode of UDP message
The 7th step: background server is accepted the UDP message from measuring appliance, and reads heading details and time stab information from the UDP message,
The 8th step: according to the sourcesink IP address of heading, and the address space of each measurement point that disposes in server administration, heading information and time stab information thereof are stored among the intermediate storage result space packet_storage in path under this message,
The 9th step: continue to handle the next message that arrives according to the 7th step, if Measuring Time enters a new time granularity interval, then heading information and the time stab information of putting in the last time granularity interval the intermediate storage result space packet_storage in local measurement point path from cooperative in the cooperative point thereof is sent in the local measurement point server, sampling proportion with cooperative point is sent to the local measurement point simultaneously
The 10th step: compare the sampling proportion of local measurement point and the sampling proportion of cooperative point, use little sampling proportion to continue to use 5.2 methods that went on foot for 5.5 steps to carry out secondary sample to the data in the last time grain size intervals of intermediate storage result space with big sampling proportion
The 11st step: cooperative is put in the local measurement point path in the last time grain size intervals of cooperative point in the intermediate storage result space number number1 of element deduct local measurement and put in the time grain size intervals number number2 of element in the intermediate storage result space, and with the two result divided by the number number1 of element in the intermediate storage result space in the last time grain size intervals of cooperative point, its result is exactly the unidirectional packet loss from the cooperative point to local measurement point path, and the result is stored among the final measurement memory space performance_storage in this path
The 12nd step: relatively cooperative is put in the local measurement point path in the last time grain size intervals of cooperative point the element in the intermediate storage result space and local measurement and is put the element in the intermediate storage result space in the time grain size intervals, select the source IP of element in two set, place IP, source port, the place port, the heading data that message identification IPID is identical, and the element timestamp that uses cooperative point deducts the element time stamp data of local measurement point, and the result is stored in the set of time difference
The 13rd step: from the time difference set, read the number of set element, the value of least member and the value of middle big or small element, value with least member is minimum one-way latency, value with the big or small element in centre is the intermediate value one-way latency, and with above-mentioned three values store into from the cooperative point to local measurement point path final measurement memory space performance_storage
The 14th step: the result of the number of minimum one-way latency, intermediate value one-way latency, unidirectional packet loss and measurement on will going up in the time grain size intervals from the cooperative point to local measurement point path is sent to the final measurement memory space performance_storage from the cooperative point to local measurement point path of cooperative point server.
Embodiment 2
The 1st step: on network, arrange at least 2 monitoring points, measuring appliance and server are set on each monitoring point, dispose the monitoring range of each monitoring point in each monitoring point, whole system can be measured the performance of n (n-1) paths at most, for a measurement point, for the measuring route of putting cooperative point from local measurement with from the measuring route that cooperative is put local measurement point two memory spaces are set all, a memory space is used to store intermediate object program packet_storage, another memory space is used to store the final measurement performance_storage in this path, at each monitoring point GP configuring S clock system, tell the light of 10% light intensity to measuring appliance by optical splitter from measuring optical fiber in each monitoring point, for example: two measurement points are set: measurement point A and measurement point B, (A-＞B) and measurement point B are to measurement point A (two paths of B-＞A) in measurement point A monitoring path from measurement point A to measurement point B, same measurement point B also monitors path from measurement point A to measurement point B, and (A-＞B) and measurement point B be (two paths of B-＞A) to the path of measurement point A, two memory spaces are set on each path
The 2nd step: identical key value key of configuration on each measurement point, two Buffer Pools are set: storage Buffer Pool basket1 and intransit buffering pond basket2 on measuring appliance, and the time granularity time_scale (is 5 minutes as the time_scale granularity is set) of performance processing
The 3rd step: configuration sampling proportion ratio separately on each measurement point, sampling proportion ratio span is between 0 to 1, and a sampling mask lengths sets mask_length is set, and the ensemble space size is 16, initial value is empty, with sampling proportion ratio successively with 1/2
1, 1/2
2, 1/2
3, 1/2
4, 1/2
5, 1/2
6, 1/2
7, 1/2
8, 1/2
9, 1/2
10, 1/2
11, 1/2
12, 1/2
13, 1/2
14, 1/2
15, 1/2
16This 16 number compares, if ratio is greater than a certain number, then with first less than the denominator index recording of the number of sampling proportion ratio in sampling mask lengths sets mask_length, sampling proportion ratio deduct this first continue to compare less than the result of the number back gained of sampling proportion ratio with after this number, with in the number after this first less than the denominator index recording of gained result's number in sampling mask lengths sets mask_length, then the gained result is deducted above-mentioned first number less than the gained result, so repeatedly, till this 16 number is relatively finished
Algorithm: establishing sampling mask length, to deposit array be mask_length[]=" ";
Sampling mask number n umber=0;
Sampling ratio is ratio;
for i from 1 to 16
{
if ratio＞＝1/2
i then
ratio＝ratio-1/2
i
mask_length[number]＝i
number++；
end if
}
return number，mask_length[]
For example:
Resolve into following formula with 0.356 and 0.41:
0.356＝1/2
2+1/2
4+1/2
5+1/2
7+1/2
8+1/2
11+1/2
15+0.00001245
0.41＝1/2
2+1/2
3+1/2
5+1/2
9+1/2
10+1/2
11+1/2
12+1/2
14+1/2
16+0.000011
The 4th step: from sampling mask lengths sets mask_length, read element successively, define a Bit String, the length of this Bit String is the current value that reads element, with in the mask_length set less than being set to 0 on the current all elements value position of reading element value, be set to 1 on all the other positions, this Bit String as a mask, is recorded among the mask set mask
For example:
0.356 coupling mask table
| The mask position | Sampling ratio (mark) | Sampling ratio (decimal) | Sampling mask |
| 2 | 1/2 2 | 0.25 | 11 |
| 4 | 1/2 4 | 0.125 | 1101 |
| 5 | 1/2 5 | 0.0625 | 10101 |
| 7 | 1/2 7 | 0.0078125 | 1100101 |
| 8 | 1/2 8 | 0.00390625 | 10100101 |
| 11 | 1/2 11 | 0.00048828125 | 11100100101 |
| 15 | 1/2 15 | 0.000030517578125 | 111101100100101 |
0.41 coupling mask table
| The mask position | Sampling ratio (mark) | Sampling ratio (decimal) | Sampling mask |
| 2 | 1/2 2 | 0.25 | 11 |
| 3 | 1/2 3 | 0.125 | 101 |
| 5 | 1/2 5 | 0.03125 | 11001 |
| 9 | 1/2 9 | 0.001953125 | 111101001 |
| 10 | 1/2 10 | 0.0009765625 | 1011101001 |
| 11 | 1/2 11 | 0.00048828125 | 10011101001 |
| 12 | 1/2 12 | 0.000244140625 | 100011101001 |
| 14 | 1/2 14 | 0.00006103515625 | 11000011101001 |
| 16 | 1/2 16 | 0.0000152587890625 | 1101000011101001 |
Can know that from 0.356 sampling mask table and 0.41 sampling mask table second mask 101 in the 0.41 sampling mask table comprises second to the 6th content of 0.356 sampling mask table.
The 5th step: measuring appliance is received each message execution following steps:
5.1 in a single day measuring appliance receives that a message makes a call to a current time stamp of measuring appliance for this message immediately;
5.2 from heading, extract source IP, the place IP of 32 bits of 32 bits, the source port of 16 bits, the place port of 16 bits, the message identification IPID of 16 bits,
5.3 back 16 bits of source IP are carried out 3 left loop computations, and 16 bits carry out XOR behind result that will generate and the place IP, generate one 16 Bit String hash1; Preceding 16 bits of source IP are carried out 4 left loop computations, and the result that will generate and preceding 16 bits of place IP carry out XOR, generate one 16 Bit String hash2; 16 bits of source port are carried out 5 left loop computations, and 16 bits of result that will generate and place port carry out XOR, generate one 16 Bit String hash3; With above-mentioned 3 16 Bit String hash1, hash2 and hash3 and message identification IPID carry out XOR, generate the cryptographic Hash hash of 16 Bit Strings,
Algorithm:
If bsip preceding 16 bits that are source IP, asip are that 16 bits behind the IP of source, bdip are that preceding 16 bits of place IP, adip are that 16 bits behind the IP of place, sport are that source port, dport are that place port, IPID are message identification;
If hash1, hash are 16 bit unsigned ints;
hash＝IPID；
hash1＝asip<<3|asip>>(16-3)；
hash1＝hash1^adip；
hash^＝hash1；
hash1＝bsip<<4|bsip>>(16-4)；
hash1＝hash1^bdip；
hash^＝hash1；
hash1＝sport<<5|sport>>(16-5)；
hash1＝hash1^dport；
hash^＝hash1；
5.4 the cryptographic Hash hash and the key key that generate are carried out the final cryptographic Hash hash of XOR generation;
Algorithm:
hash＝hash^key
5.5 from mask set mask, read element successively, first bit extracts the Bit String of mask-length since the final hash value hash of 16 bits, and this Bit String and the current mask that reads compared, extraction has the message of the Bit String identical with mask, and 8 byte times of preceding 40 bytes of this heading and this message are stabbed be stored in the current Buffer Pool
Algorithm:
Each sampling mask lengths sets is mask_length[i], (i=1, number), wherein number represents the number of mask;
Cur mask=0; First sub-mask of // appointment;
For (i==1; I＜=16; I++) // ident value has 16 bits, so maximum the needs circulates 16 times;
If (hash[i]==0) //hash[i] value (0 or 1) of i bit of expression ident value;
{
If (i＞mask_length[cur_mask]) this message of sampling;
Else if (i＜mask_length[cur_mask]) this message of unsample;
-11-
else if(i＝＝mask_length[cur_mask])
{
If (cur_mask==number) // has pointed to last mask
This message of unsample
else
Cur mask++; // search next sampling mask
}
}
The 6th step: continued to carry out the 5th step and handle newly arrived message, if current Buffer Pool is full, current Buffer Pool is set to the intransit buffering pond, another Buffer Pool in the measuring appliance is set to current Buffer Pool simultaneously, and the data in the intransit buffering pond are transmitted to background server in the mode of UDP message
The 7th step: background server is accepted the UDP message from measuring appliance, and reads heading details and time stab information from the UDP message;
The 8th step: according to the sourcesink IP address of heading, and the address space of each measurement point that disposes in server administration, heading information and time stab information thereof are stored among the intermediate storage result space packet_storage in path under this message,
The 9th step: continue to handle the next message that arrives according to the 7th step, if Measuring Time enters a new time granularity interval, then heading information and the time stab information of putting in the last time granularity interval the intermediate storage result space packet_storage in local measurement point path from cooperative in the cooperative point thereof is sent in the local measurement point server, sampling proportion with cooperative point is sent to the local measurement point simultaneously
The 10th step: compare the sampling proportion of local measurement point and the sampling proportion of cooperative point, use little sampling proportion to continue to use 5.2 methods that went on foot for 5.5 steps to carry out secondary sample to the data in the last time grain size intervals of intermediate storage result space with big sampling proportion
The 11st step: cooperative is put in the local measurement point path in the last time grain size intervals of cooperative point in the intermediate storage result space number number1 of element deduct local measurement and put in the time grain size intervals number number2 of element in the intermediate storage result space, and with the two result divided by the number number1 of element in the intermediate storage result space in the last time grain size intervals of cooperative point, its result is exactly the unidirectional packet loss from the cooperative point to local measurement point path, and the result is stored among the final measurement memory space performance_storage in this path
The 12nd step: relatively cooperative is put in the local measurement point path in the last time grain size intervals of cooperative point the element in the intermediate storage result space and local measurement and is put the element in the intermediate storage result space in the time grain size intervals, select the source IP of element in two set, place IP, source port, the place port, the heading data that message identification IPID is identical, and the element timestamp that uses cooperative point deducts the element time stamp data of local measurement point, and the result is stored in the set of time difference
The 13rd step: the value (intermediate value one-way latency) that from the time difference set, reads the number of set element, the value of least member (minimum one-way latency) and middle big or small element, and with these three values store into from the cooperative point to local measurement point path final measurement memory space performance_storage
The 14th step: the result of the number of minimum one-way latency, intermediate value one-way latency, unidirectional packet loss and measurement on will going up in the time grain size intervals from the cooperative point to local measurement point path is sent to the final measurement memory space performance_storage from the cooperative point to local measurement point path of cooperative point server.
Claims (2)
1, a kind of terminal to terminal running performance monitoring method based on sampling measurement is characterized in that:
The 1st step: on network, arrange at least 2 monitoring points, measuring appliance and server are set on each monitoring point, dispose the monitoring range of each monitoring point in each monitoring point, whole system can be measured the performance of n (n-1) paths at most, for a measurement point, for the measuring route of putting cooperative point from local measurement with from the measuring route that cooperative is put local measurement point two memory spaces are set all, a memory space is used to store intermediate object program packet_storage, another memory space is used to store the final measurement performance_storage in this path, at each monitoring point GP configuring S clock system, tell the light of 10% light intensity to measuring appliance by optical splitter from measuring optical fiber in each monitoring point
The 2nd step: identical key value key of configuration on each measurement point is provided with two Buffer Pools on measuring appliance: storage Buffer Pool basketl and intransit buffering pond basket2, and the time granularity time_scale of performance processing,
The 3rd step: configuration sampling proportion ratio separately on each measurement point, sampling proportion ratio span is between 0 to 1, and a sampling mask lengths sets mask_length is set, and the ensemble space size is 16, initial value is empty, with sampling proportion ratio successively with 1/2
1, 1/2
2, 1/2
3, 1/2
4, 1/2
5, 1/2
6, 1/2
7, 1/2
8, 1/2
9, 1/2
10, 1/2
11, 1/2
12, 1/2
13, 1/2
14, 1/2
15, 1/2
16This 16 number compares, if ratio is greater than a certain number, then with first less than the denominator index recording of the number of sampling proportion ratio in sampling mask lengths sets mask_length, sampling proportion ratio deduct this first continue to compare less than the result of the number back gained of sampling proportion ratio with after this number, with in the number after this first less than the denominator index recording of gained result's number in sampling mask lengths sets mask_length, then the gained result is deducted above-mentioned first number less than the gained result, so repeatedly, till this 16 number is relatively finished
The 4th step: from sampling mask lengths sets mask_length, read element successively, define a Bit String, the length of this Bit String is the current value that reads element, with in the mask_length set less than being set to 0 on the current all elements value position of reading element value, be set to 1 on all the other positions, this Bit String as a mask, is recorded among the mask set mask
The 5th step: measuring appliance is received each message execution following steps:
5.1 in a single day measuring appliance receives that a message makes a call to a current time stamp of measuring appliance for this message immediately;
5.2 from heading, extract source IP, the place IP of 32 bits of 32 bits, the source port of 16 bits, the place port of 16 bits, the message identification IPID of 16 bits,
5.3 back 16 bits of source IP are carried out 3 left loop computations, and 16 bits carry out XOR behind result that will generate and the place IP, generate one 16 Bit String hash1; Preceding 16 bits of source IP are carried out 4 left loop computations, and the result that will generate and preceding 16 bits of place IP carry out XOR, generate one 16 Bit String hash2; 16 bits of source port are carried out 5 left loop computations, and 16 bits of result that will generate and place port carry out XOR, generate one 16 Bit String hash3; With above-mentioned 3 16 Bit String hash1, hash2 and hash3 and message identification IPID carry out XOR, generate the cryptographic Hash hash of 16 Bit Strings,
5.4 the cryptographic Hash hash and the key key that generate are carried out the final cryptographic Hash hash of XOR generation;
5.5 from mask set mask, read element successively, first bit extracts the Bit String of mask-length since the final hash value hash of 16 bits, and this Bit String and the current mask that reads compared, extraction has the message of the Bit String identical with mask, and 8 byte times of preceding 40 bytes of this heading and this message are stabbed be stored in the current Buffer Pool
The 6th step: continued to carry out the 5th step and handle newly arrived message, if current Buffer Pool is full, current Buffer Pool is set to the intransit buffering pond, another Buffer Pool in the measuring appliance is set to current Buffer Pool simultaneously, and the data in the intransit buffering pond are transmitted to background server in the mode of UDP message
The 7th step: background server is accepted the UDP message from measuring appliance, and reads heading details and time stab information from the UDP message,
The 8th step: according to the sourcesink IP address of heading, and the address space of each measurement point that disposes in server administration, heading information and time stab information thereof are stored among the intermediate storage result space packet_storage in path under this message,
The 9th step: continue to handle the next message that arrives according to the 7th step, if Measuring Time enters a new time granularity interval, then heading information and the time stab information of putting in the last time granularity interval the intermediate storage result space packe_storage in local measurement point path from cooperative in the cooperative point thereof is sent in the local measurement point server, sampling proportion with cooperative point is sent to the local measurement point simultaneously
The 10th step: compare the sampling proportion of local measurement point and the sampling proportion of cooperative point, use little sampling proportion to continue to use 5.2 methods that went on foot for 5.5 steps to carry out secondary sample to the data in the last time grain size intervals of intermediate storage result space with big sampling proportion
The 11st step: cooperative is put in the local measurement point path in the last time grain size intervals of cooperative point in the intermediate storage result space number number1 of element deduct local measurement and put in the time grain size intervals number number2 of element in the intermediate storage result space, and with the two result divided by the number number1 of element in the intermediate storage result space in the last time grain size intervals of cooperative point, its result is exactly the unidirectional packet loss from the cooperative point to local measurement point path, and the result is stored among the final measurement memory space performance_storage in this path
The 12nd step: relatively cooperative is put in the local measurement point path in the last time grain size intervals of cooperative point the element in the intermediate storage result space and local measurement and is put the element in the intermediate storage result space in the time grain size intervals, select the source IP of element in two set, place IP, source port, the place port, the heading data that message identification IPID is identical, and the element timestamp that uses cooperative point deducts the element time stamp data of local measurement point, and the result is stored in the set of time difference
The 13rd step: from the time difference set, read the number of set element, the value of least member and the value of middle big or small element, value with least member is minimum one-way latency, value with the big or small element in centre is the intermediate value one-way latency, and with above-mentioned three values store into from the cooperative point to local measurement point path final measurement memory space performance_storage
The 14th step: the result of the number of minimum one-way latency, intermediate value one-way latency, unidirectional packet loss and measurement on will going up in the time grain size intervals from the cooperative point to local measurement point path is sent to the final measurement memory space performance_storage from the cooperative point to local measurement point path of cooperative point server.
2, the terminal to terminal running performance monitoring method based on sampling measurement according to claim 1 is characterized in that the time granularity time_scale that described performance of the 2nd step is handled is set to 5 minutes.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB2005100376650A CN100361461C (en) | 2005-01-11 | 2005-01-11 | Terminal to terminal running performance monitoring method based on sampling measurement |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB2005100376650A CN100361461C (en) | 2005-01-11 | 2005-01-11 | Terminal to terminal running performance monitoring method based on sampling measurement |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN1645825A CN1645825A (en) | 2005-07-27 |
| CN100361461C true CN100361461C (en) | 2008-01-09 |
Family
ID=34876167
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNB2005100376650A Expired - Fee Related CN100361461C (en) | 2005-01-11 | 2005-01-11 | Terminal to terminal running performance monitoring method based on sampling measurement |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN100361461C (en) |
Families Citing this family (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101212358B (en) * | 2006-12-28 | 2010-05-19 | 中国移动通信集团公司 | Network performance measuring method |
| CN101212361B (en) * | 2006-12-28 | 2011-06-15 | 中国移动通信集团公司 | Network performance measuring method |
| CN101212360B (en) * | 2006-12-28 | 2010-12-29 | 中国移动通信集团公司 | Network performance measuring method |
| US20080181134A1 (en) * | 2007-01-29 | 2008-07-31 | Nikolaos Anerousis | System and method for monitoring large-scale distribution networks by data sampling |
| EP2556623A4 (en) * | 2010-03-24 | 2017-03-08 | Nokia Technologies Oy | Method and apparatus for establishing an electrical connection and establishing a fiber optic connection |
| CN101938459A (en) * | 2010-06-22 | 2011-01-05 | 北京豪讯美通科技有限公司 | CRNET (China Railcom Net) sSafe cooperative defense system for whole course communication network |
| CN106845840A (en) * | 2016-09-29 | 2017-06-13 | 成都市食品药品检验研究院 | Performance appraisal based on gps data application, real-time early warning system |
| CN107465771B (en) * | 2017-09-20 | 2020-05-19 | 广东工业大学 | Network address segment dividing method and system |
| CN107589991A (en) * | 2017-10-31 | 2018-01-16 | 麦格创科技（深圳）有限公司 | The webpage distribution method and system of distributed system |
| CN110943877B (en) * | 2018-09-21 | 2022-02-22 | 华为技术有限公司 | Network state measuring method, equipment and system |
| IT201800010791A1 (en) | 2018-12-04 | 2020-06-04 | Telecom Italia Spa | Performance measurement in a packet-switched communications network |
| CN114143086B (en) * | 2021-11-30 | 2023-09-26 | 北京天融信网络安全技术有限公司 | Web application identification method and device, electronic equipment and storage medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6421349B1 (en) * | 1997-07-11 | 2002-07-16 | Telecommunications Research Laboratories | Distributed preconfiguration of spare capacity in closed paths for network restoration |
| US6738355B1 (en) * | 2000-11-01 | 2004-05-18 | Agilent Technologies, Inc. | Synchronization method for multi-probe communications network monitoring |
-
2005
- 2005-01-11 CN CNB2005100376650A patent/CN100361461C/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6421349B1 (en) * | 1997-07-11 | 2002-07-16 | Telecommunications Research Laboratories | Distributed preconfiguration of spare capacity in closed paths for network restoration |
| US6738355B1 (en) * | 2000-11-01 | 2004-05-18 | Agilent Technologies, Inc. | Synchronization method for multi-probe communications network monitoring |
Non-Patent Citations (3)
| Title |
|---|
| Distributed Sampling Measurement Model in a Large-ScaleHigh-Speed IP Networks. GongJian, ChengGuang.Journal of Southeast University,Vol.18 No.1. 2002 * |
| 基于分组标识的网络流量抽样测量模型. 程光，龚俭，丁伟.电子学报，第30卷第12A期. 2002 * |
| 基于统计分析的高速网络分布式抽样测量模型. 程光，龚俭，丁伟.计算机学报，第26卷第10期. 2003 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN1645825A (en) | 2005-07-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN100361461C (en) | Terminal to terminal running performance monitoring method based on sampling measurement | |
| CN111602370B (en) | In-band telemetry using limited extra bytes | |
| US8045478B2 (en) | Performance measurement in a packet transmission network | |
| US20090180393A1 (en) | Sampling apparatus distinguishing a failure in a network even by using a single sampling and a method therefor | |
| US20150271034A1 (en) | System and method for accurately displaying communications traffic information | |
| US7898969B2 (en) | Performance measurement in a packet transmission network | |
| CN103401716B (en) | The method of network performance measurement, Network Management Equipment and network element device | |
| Plonka et al. | Assessing performance of Internet services on IPv6 | |
| Kocak et al. | Performance measurement of IP networks using two-way active measurement protocol | |
| Brownlee | Flow-based measurement: IPFIX development and deployment | |
| CN107454647A (en) | The sending method and device of periodicity routing iinformation | |
| CN110838950B (en) | Method and device for determining network performance jitter value | |
| CN102124698B (en) | System and method for exporting structured data in a network management environment | |
| Corral et al. | End-to-end active measurement architecture in ip networks (saturne) | |
| Aceto et al. | Open source platforms for Internet Monitoring and Measurement | |
| JP4938042B2 (en) | Flow information transmitting apparatus, intermediate apparatus, flow information transmitting method and program | |
| Wang et al. | A new virtual indexing method for measuring host connection degrees | |
| CN112737874B (en) | Packet loss statistics realization method and device for message | |
| Ehrlich et al. | Passive flow monitoring of hybrid network connections regarding quality of service parameters for the industrial automation | |
| Liu et al. | SFANT: A SRv6-Based Flexible and Active Network Telemetry Scheme in Programming Data Plane | |
| Claffy et al. | The 10th workshop on active internet measurements (AIMS-10) report | |
| Zhang et al. | INT-Balance: In-Band Network-Wide Telemetry with Balanced Monitoring Path Planning | |
| Graur et al. | Steps towards decentralized deterministic network coding | |
| Shi et al. | Intrinsic monitoring within an IPv6 network: mapping node information to network paths | |
| Kobayashi et al. | IP Multicast Traffic Measurement Method with IPFIX/PSAMP |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| C17 | Cessation of patent right | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20080109 Termination date: 20110111 |